Bug #119004 UNREACHABLE member is not expelled after the group_replication_member_expel_timeout delay
Submitted: 12 Sep 16:07 Modified: 11 Oct 3:23
Reporter: Garrido Mickael Email Updates:
Status: Open Impact on me:
None 
Category:MySQL Server Severity:S2 (Serious)
Version:8.0+ OS:Any
Assigned to: CPU Architecture:Any

[12 Sep 16:07] Garrido Mickael 
Description:
We're performing a MySQL upgrade from 8.0.x to 8.0.43 to several clusters. During the process, some members occasionally become UNREACHABLE, and in the logs of other members, we see messages like: 'Old incarnation found while trying to add node ...'.
According to the documentation, a member marked as UNREACHABLE should be evicted after the delay defined by the "group_replication_member_expel_timeout" variable. However, this never happens.
We use InnoDB Cluster with Group Replication, which relies on XCom for internal communication.
Are you aware of any issues related to this?
We’ve tried using removeInstance followed by addInstance or rejoinInstance, but none of those approaches worked. The only way we’ve been able to recover is by restarting the entire cluster.

How to repeat:
Create a cluster and upgrade to a minor version.
[15 Sep 14:02] MySQL Verification Team 
Hi,

I just did upgrade few days ago of a 5 node innodb cluster from 8.0.35 to 8.0.43 without a glitch.

Can you share more data? (configuration, logs ..)

Thanks